Chuluun-Qwen2.5-72B-v0.01
GGUF quants available here: https://huggingface.co/bartowski/Chuluun-Qwen2.5-72B-v0.01-GGUF
This is a merge of pre-trained language models created using mergekit.
The models in this merge are some of my favorites and I found I liked all of them for different reasons. I believe this model is greater than the sum of its parts - it has the storywriting and language of Eva and Kunou, the spiciness of Magnum, and the uncensored intelligence of Tess. It excels in handling multiple characters and keeping their thoughts, speech, and actions separate, including scene changes. It also appears to match dialogue well to the characters and their backgrounds.
Model_stock was the method used, it's very straightforward and quite fast, the bottleneck seemed to be my NVMe drive.
All source models use ChatML prompt formatting and it responds very well. Consider the following settings (thanks Geechan!): Temp 1.25, MinP 0.02, XTC 0.15/probability 0.5, DRY 0.8. All other samplers neutralized. Chuluun seems to be able to work with higher temperatures than other Qwen models without losing coherency.
Konnect has released their Qwenception sysprompts and settings, which work quite well with Chuluun.
Chuluun is built off of uncensored models and testing has confirmed it will go to dark places. Your use of this model assumes you to understand this and that you're responsible for what you do with it.
Merge Details
Merge Method
This model was merged using the Model Stock merge method using migtissera/Tess-v2.5.2-Qwen2-72B as a base.
Models Merged
The following models were included in the merge:
- EVA-UNIT-01/EVA-Qwen2.5-72B-v0.2
- Sao10K/72B-Qwen2.5-Kunou-v1
- anthracite-org/magnum-v4-72b
Configuration
The following YAML configuration was used to produce this model:
models:
- model: EVA-UNIT-01/EVA-Qwen2.5-72B-v0.2
- model: Sao10K/72B-Qwen2.5-Kunou-v1
- model: anthracite-org/magnum-v4-72b
merge_method: model_stock
base_model: migtissera/Tess-v2.5.2-Qwen2-72B
parameters:
filter_wise: false
dytpe: float16
name: DatToad/Chuluun-Qwen2.5-72B-v0.01
Thank Yous!
My work is built on the backs of giants - all I did was some cooking in the kitchen. Much credit goes to all of the following:
-
Allura-Org, for the EVA models and their kind support as I've learned some of the finer points of working with LLMs.
-
Sao10k, creator of Euryale and Kunou and inspiring so many writers along the way
-
Sophosympatheia, their original merge of the legendary Midnight Miqu has entertained countless writers and inspired me to give merging a try
-
#horde in the KoboldAI Discord, who've also answered a lot of questions I've had